Mobilenet and Deep Residual Network for Object Detection and Classification of Objects in IoT Enabled Construction Safety Monitoring

Construction areas are extremely dangerous since the dynamic communication amongst workers and moving tools along with wide fatality rates occurred by collision. Falls, struck-bys, and caught-in-between are the general kinds of accidents that happen in work area. In spite of having their unique benefits, more number of day–to-day accident prevention events respond passively to situations where workers or tools enter predefined unsafe areas. To prevent these types of accidents, a new scheme is introduced for object detection (OD) and classification of objects in IoT-enabled constructing safety monitoring named Mobile-Residual Network (MR-Net), where MR-Net is merged by MobileNet and Deep Residual Network (DRN). This module is composed of three sub-systems, prevention of backover accidents, smart alerting for probable risk avoidance, and middleware IoT platform. The first technique has IoT network with sensor unit. The second technique has cloud server employed to provide smart alerts. The third one is detection progress carried out by image processing with the aid of collected images. First, IoT nodes are simulated, wherein IoT image sensors are responsible for collecting data. At base station, the collected data is pre-progressed by utilizing Non- Local Means (NLM) filter. After that, object is detected by you only live once v3 (YOLO v3) where the recognized objects, like person, machines, vehicles and safety cones are detected. If the detected object is person then classify the objects like Hardhat, Mask and Safety Vest worn using MR-Net. The analytic measures utilized for MR-Net such as accuracy, True Positive Rate (TPR), True Negative Rate (TNR), Positive predictive value (PPV) and Negative predictive value (NPV) gained maximal values of 0.912, 0.905, 0.923, 0.909 and 0.919.